Feature-Based Simultaneous Localization and Mapping for LiDAR-Equipped Lunar Surface Mobility Systems Using Random Finite Set Filtering
Tuan Dinh Luong, Jordan D. Larson
University of Alabama
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
There is a renewed international effort to establish a sustained human presence on the lunar surface, e.g., the confirmed existence of water ice at the lunar poles has rendered the exploration of these regions particularly compelling. Current proposed missions predominantly employ robotic rovers, the majority of which operate in low-latitude environments. In contrast, high-latitude exploration is impeded by challenging illumination conditions, which NASA is currently mitigating through the deployment of high-power illuminators and passive imaging systems. This work investigates alternative sensing modalities independent of the lunar surface’s extreme lighting conditions. Specifically, it evaluates the use of scanning-mirror light detection and ranging (LiDAR) for terrain-relative navigation, with the objective of reducing drift errors inherent to dead-reckoning techniques. In addition, it examines simultaneous localization and mapping (SLAM) strategies that incorporate multiobject tracking frameworks to address measurement uncertainty, missed detections, environmental clutter, and unknown data associations. To this end, several extensions of the probability hypothesis density (PHD) SLAM algorithm are developed to adapt it for the lunar rover application. These extensions include 3-D state modeling, the use of ground-affine landmarks, explicit clutter representation, and the decoupling of odometry and exteroceptive sensor frames. The extended algorithm is integrated into a comprehensive localization pipeline encompassing high-fidelity sensor modeling and pose estimation. Simulation studies demonstrate that PHD-SLAM consistently reduces localization error relative to odometry alone. Furthermore, PHD-SLAM achieves accurate pose estimation with noisy limited field-of-view LiDAR measurements in scenarios where point cloud alignment techniques, such as iterative closest point, fail to converge.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
计算机 / AIRobotics and Sensor-Based Localization
Target Tracking and Data Fusion in Sensor Networks · Inertial Sensor and Navigation